Intelligence Raises $7.9 Million as Design Arena Expands AI Model Evaluation

Artificial intelligence startup Intelligence has emerged with $7.9 million in seed funding to expand Design Arena, a platform that uses large-scale human feedback to help improve AI-generated content. The round was led by Index Ventures, with participation from Conviction, A*, Valkyrie and other investors.

The company traces its origins to 2025, when co-founder and CEO Grace Li and several college friends were building an AI-powered game engine shortly before graduation. Although the models they created could generate functional games, the team found there was no reliable way to determine whether those games were actually enjoyable. They concluded that human judgment remained essential, prompting them to build a system capable of collecting feedback at scale.

That idea quickly found commercial demand. According to Li, AI companies developing frontier models were looking for reliable ways to evaluate creative outputs, and Intelligence secured its first major customer shortly after launching.

Ranking AI-generated content

For individual users, Design Arena functions like an advanced AI model comparison platform. Users submit prompts through a ChatGPT-style interface, choose a format such as websites, images or other visual content, and receive multiple AI-generated results. Instead of accepting a single answer, users rank competing outputs through a series of head-to-head comparisons until the platform produces an overall ranking.

Those comparisons generate valuable evaluation data for enterprise customers. Because users generally focus on selecting the best output rather than supporting a particular AI model, their preferences provide model developers with continuous feedback about what people actually prefer.

Li said Design Arena has grown to 5.3 million users worldwide and is generating $60 million in annual recurring revenue. The platform also requires users to log in before receiving results, allowing Intelligence to analyse how design preferences vary across countries and evolve. Li noted, for example, that users in Asia often prefer more visually dense dashboard designs.

The company believes those human preferences complement automated AI benchmarks, which can operate at larger scale but may be vulnerable to manipulation. Li pointed to the recent Hugging Face security breach as an example of why dependable evaluation methods remain important.

Growing demand for human evaluation

The market for AI evaluation remains competitive. Startup Yupp shut down earlier this year despite raising $33 million, attracting more than 1.3 million users and signing several frontier AI companies as customers.

Other businesses have continued to attract investor interest. LM Arena, which applies a similar ranking system to text-based AI responses, raised $150 million in a Series A funding round in January, only a few months after launching its paid product.

For Intelligence, the latest funding will support further development of Design Arena as AI companies continue searching for scalable ways to incorporate human preferences into model improvement.